What does the evidence say about the value of the hospitalist model?

A while back I suggested that it was unlikely there would be any more studies on the effects of the hospitalist model on outcomes and resource utilization because there were no longer enough practitioners in the traditional model of hospital care against which to compare. I also suggested that despite the hype, good quality evidence in favor of the value of the model was lacking. Although I stand by that premise the latest systematic review, published in Mayo Clinic Proceedings, warrants a revisit to the issue.

The review concluded, in line with popular belief, that the hospitalist model results in improved efficiency of resource utilization. Although it will be greeted with fanfare among boosters of the model (read Happy Hospitalist’s take here) there is a big problem with this paper which the author hints at (italics mine):

Systematic reviews may be hampered by difficulties related to publication bias, in which articles are more likely to be published if they show positive findings. This limitation is not confined to this review but is a potential problem for any review. I am unaware of any unpublished data on the topic of this review. Whether to include unpublished data should be an important consideration in conducting a systematic review. Investigators need to remember, however, that bias against negative results is not the only reason why a manuscript may be unpublished; a manuscript may have any of a number of inadequacies that disqualify it from consideration for publication.

It turns out that there are very important unpublished data on this topic, and those data are not friendly to supporters of the movement. A very large and methodologically sound study was presented at the 2005 national meeting of the Society of Hospital Medicine. When it first came out it received a lot of attention in the blogs. DB said of the study:

Much of the fire behind the rapid growth of the hospitalist movement comes early studies which suggested that hospitalists made a large difference in cost and quality of care. This study suggests that we must be careful in attributing benefits. I do believe that there is a valid volume efficacy curve, but the big question is how we determine it. Having the title does not make one a better inpatient doctor.

California Medicine man observed:

Great. It was for my expertise as a hospitalist that I received my academic appointment at UCLA. I can imagine my department head at the next budget meeting. 'Now remind me, why did we hire Ford?'In fact, I'm giving a grand rounds lecture on the benefits of the hospitalist paradigm in a few weeks. I guess I can make it a very brief one. I'm reminded of Gilda Radner's line while playing character Emily Litella: "Never mind!"

The study was never published in any Medline indexed journal, which is why it was soon forgotten and didn’t make it into the systematic review. (There just might be a lesson here on the effect of publication bias from within our own ranks). The presentation abstract, though, was reported in the summer 2005 issue of The Hospitalist which you can access here. As far as I can tell it was the highest quality study ever done. It was a large (the second largest ever done), prospective, randomized multicenter study. It showed no difference between hospitalists and non-hospitalists in any efficiency metric or in patient outcomes.

But back to the Mayo Clinic Proceedings paper. The abstract says:

The reports that were included (N=33) show general agreement that hospitalist care leads to shorter length of stay and lower cost per stay.

This was not a meta-analysis so we can’t do sensitivity testing. It’s not difficult to imagine, though, how the conclusion might read had the 2005 SHM study been included in the review. It would have been much more circumspect and might have gone something like this:

Most studies comparing models of care are of low methodologic quality. Although evidence from these studies suggests reduced length of stay and charges per case attributable to the hospitalist model the only large randomized, prospective multicenter study showed no significant difference between models in any efficiency metric.

So if the evidence doesn’t support the present day hospitalist hype what does it tell us? I think the following:

For mortality and other patient outcomes it’s value neutral. For efficiency metrics (length of stay and charge per case) at least we know it’s not detrimental. It may be beneficial but evidence is mixed and will remain controversial.

What’s the harm in hyping the evidence? For one thing it may lead to inflated economic expectations of hospitalist groups, causing administrators to view hospitalists as business solutions. Back in the early days of my hospitalist program there was an excellent, tightly run traditional internal medicine group which admitted patients to our hospital. We compared efficiency metrics but try as we might we could not beat theirs. In measured value we offered nothing over them. Fortunately, in spite of that fact, demand for our services grew rapidly and our niche was secured. But with today’s economic woes and all the hospitalists-as-business-solutions hype I can imagine administration saying “all the studies show that hospitalists can deliver more efficient care, so why aren’t you guys more efficient than your local non-hospitalist peers?” (Now as it turns out horrible reimbursement and deteriorating professional satisfaction in general internal medicine led many of those excellent docs to look for greener pastures, often in hospital medicine, causing that internal medicine group to dissolve).

So what, other than the need to fill a growing niche, is the value of hospitalists? For me it’s largely subjective. My group members and I feel good about the work we do. Our physician colleagues in the community appreciate it, as do our hospital staff and administration. But the economic benefits, if any, of hospitalist care vary from one community to another. If your group can make a difference in your hospital’s bottom line, great. Measure it. But I’m afraid we’ll never be able to make a general evidentiary claim for the hospitalist model over any other.